Model-free classification of X-ray scattering signals applied to image segmentation

Area of Science:

• Materials Science
• Biophysics
• Analytical Chemistry

Background:

• Analysis of small-angle and wide-angle X-ray scattering (SAXS/WAXS) typically requires theoretical models, which complicates interpretation for heterogeneous samples.
• Analyzing large datasets from time-resolved or scanning SAXS/WAXS methods remains challenging due to model dependency.

Purpose of the Study:

• To develop and demonstrate a model-free classification method for segmenting SAXS/WAXS signals.
• To enable the analysis of complex heterogeneous samples without prior knowledge or theoretical models.
• To facilitate the interpretation of scattering data from scanning SAXS/WAXS maps.

Main Methods:

• A model-free classification approach based on signal inflection points is introduced.
• Dimensionality reduction and clustering algorithms are employed to classify SAXS/WAXS signals by similarity.
• The method automatically estimates the number of clusters representing distinct sample regions.

Main Results:

• The method successfully segments scanning SAXS/WAXS maps by classifying individual scattering curves.
• It allows for the extraction of representative signals from each cluster to reveal phase distribution.
• Applications demonstrated include segmentation of a mudrock sample and breast tissue lesions.

Conclusions:

• This model-free classification method offers a robust approach for analyzing complex SAXS/WAXS data.
• It simplifies the segmentation and interpretation of heterogeneous samples, particularly in scanning applications.
• The technique has broad applicability in materials science and biomedical imaging.